RealAnalyst

Use when: (1) Starting analysis to understand dataset structure, (2) Checking data quality before processing, (3) Need schema, semantic roles, null counts, and signals before analysis, (4) Need to bind a profiling result back to a specific CSV inside a continuous-analysis job. Triggers: profiling, 数据画像, 数据概览, profile, 数据质量, schema summary.

用于通用数据分析任务的总控编排与分析执行 skill。适用于监控、汇报、诊断、归因、对标、排名、探索、专项分析，以及任何需要连续对话、连续分析、单会话单 job 管理、先做需求画像、获得用户确认，再编排取数、数据体检、分析执行和报告追加流程的请求。

Controlled Tableau, DuckDB, MySQL, and ClickHouse data export for RealAnalyst/Codex. Use when an analysis task needs to fetch formal CSV data from a registered source in the unified SQLite registry, including Tableau view/domain exports with vf/vp filters and SQL table/view exports with field whitelist, filters, aggregation, and audited job metadata updates. Triggers: tableau export, duckdb export, mysql export, clickhouse export, 受控取数, 数据导出, 补数, 追加分析, export_summary, duckdb_export_summary, data_export_summary.

Use when analysis jobs, user feedback, profile outputs, or real-data probes need to be converted into reference materials for later RA:metadata maintenance; trigger for metadata 修正材料, 字段定义不清, 指标口径待修, evidence 补充, job feedback, profile-based metadata refinement.

Use when generating Markdown reports that explain RealAnalyst metadata, sync results, dataset registration details, field/metric/filter semantics, metadata gaps, or metadata inventory. Triggers: 元数据报告, 同步结果说明, 字段口径说明, 数据集注册报告, 元数据清单.

Use when: (1) Need to search for metric, field, term, dataset, or mapping definitions from the metadata index, (2) Need to browse available datasets before starting analysis, (3) Need a lightweight lookup result instead of reading full YAML files. Triggers: 搜索指标, 搜索字段, 搜索术语, 搜索数据集, metadata search, 浏览数据集目录, catalog, dataset discovery.

Use when initializing, registering, refreshing, validating, or packaging LLM-maintained dataset metadata for RealAnalyst; trigger for 数据集注册, 元数据初始化, 业务口径维护, metadata context, Tableau/DuckDB source onboarding, or analysis planning context. For search/catalog use RA:metadata-search.

Use when: (1) Starting a new analysis task and need a formal plan, (2) Need to classify the scenario and choose framework, (3) Need to write `.meta/analysis_plan.md` before analysis begins. Triggers: 分析规划, 场景识别, 框架选择, analysis plan, hypothesis-driven planning, goal decomposition.

Use when: (1) Need a specific report template from RealAnalyst references, (2) Need a planning framework configuration (logic_path, goal_template, dimension_type_hints). Triggers: 查询模板, 查询分析框架, template lookup, framework lookup. NOT for metric/field/term lookup — use RA:metadata-search instead.

Use when: (1) Need a specific template/term/metric/dimension from RealAnalyst metadata or references, (2) Need planning reference lookup, (3) Need a small lookup result instead of reading large files directly. Triggers: 查询模板, 查询术语, 查询指标定义, 查询分析框架, 查询维度定义, template lookup, glossary lookup, metric definition, framework lookup, dimension lookup.

Use when: (1) Writing the report after planning and analysis, including continuous follow-up rounds in the same job, (2) Need to execute a locked report template, (3) Need to enforce append-only report updates, timeline management, data-source, output-file-list, and appendix rules, (4) Need report naming and output decisions to follow the plan contract. Triggers: 报告撰写, report writing, 模板选择, 输出文件清单, 口径说明, 数据来源, 追加报告, 连续分析.

Use when: (1) Validating a report before delivery, (2) Need a machine-readable verification.json summary, (3) Need to confirm evidence chain, rankings, trends, appendix, user-surface, and output-file-list compliance, (4) Need to decide whether a report can pass the final gate. Triggers: verify, 验证报告, 交付前校验, verification.json, 一致性校验, 最终门禁.

Use when a user first installs RealAnalyst, asks how to start, wants to extract metadata from Tableau/DuckDB/files, or wants to know what information to prepare before registering datasets or running analysis.

Use when RealAnalyst metadata needs to be exported into a Data Analytics semantic-layer skill package for later Data Analytics analysis, source selection, metric lookup, or dashboard/report handoff. Do not use for writing Data Analytics user-context automatically, modifying Data Analytics plugin files, or changing RealAnalyst canonical metadata.

Merge multiple datasets into unified data with lineage tracking. Use when: (1) Combine datasets from different sources, (2) Union datasets with same schema, (3) Join datasets on common keys, (4) Pass through single dataset with manifest update. Triggers: fusion, 数据融合, 合并, merge, join, union, 数据合并.